Don’t Let AI Tell You How to Live
Unless you don't mind decreased autonomy and increased narcissism
Of the myriad AI applications that are touted as imminent, one of the most popular is that of the AI assistant and, in some cases, the AI therapist or life coach. This has me thinking back to an entry in Soren Kierkegaard’s journals that has stayed with me since my first semester of college:
There are many people who reach their conclusions about life like schoolboys: they cheat their master by copying the answer out of a book without having worked the sum out for themselves.
Kierkegaard is right. Many people do little thinking about the big picture of their lives. What their ideals are, how they want their lives to unfold, what counts as a life well lived. In this Substack article, a decidely less lofty venue than Kierkegaard’s journals, I’ll explore some thoughts I have about the nature of advice and the ways AI advisors may lead us astray, encourage narcissism, decrease autonomy, or all of the above. But the blame isn’t just and perhaps not even primarily on AI or its developers. The responsibility lies on us.
Seeking Advice is Well Advised, But There are Complications
Life is difficult. And I’m not just talking about grief from the loss of a loved one or disappointment or frustration or even slight terror at not getting that job, but rather the continuous effort involved in figuring out how best to steer your life, whatever comes your way. Given our limited experience, knowledge, and wisdom, it’s wise to consult people who have experience, knowledge, and wisdom that cover domains ours does not.
But there are good and bad advisors and there are good and bad advisees. I’m going to paint a picture of what a good advisor/advisee relationship looks like and then I’ll highlight three ways we can fall short of that ideal. Then I’ll turn to how the proliferation and ubiquity of AI advisors will drag people further from that ideal.
The ideal advisor/advisee relationship
Let’s put fact-finding to the side. Fact finding isn’t advice seeking. It’s an attempt to learn what you can so that you can figure out what to do. But the figuring out is often the hard part, especially when lots of facts are unknown:
Will I enjoy it?
Will this get me what I want?
Will I regret this decision?
Will I find the work fulfilling?
Will I be able to balance my career or my passions with having a (second) child?
What kind of relationship do I want to cultivate with this person?
Should I walk away from this friendship?
Do they deserve another chance?
Should I let their jealousy bother me so much?
These questions may result from a variety of sources. In some instances, it’s an absence of self-knowledge. In other more profound cases, it’s that one’s self or personality is as yet unsettled and these decisions are one that will shape or form that self.
Deliberating about whether to have a child, for instance, is not only a kind of pleasure/pain/fulfillment calculus, but also an attempt to figure out how you want your life to unfold or what sort of life you want it to be. The life of a caregiver and parent? The life of a solitary even if not lonely nomadic adventurer? Or, somewhat more blandly but still significant: the life of a doctor or the life of a researcher? The life of someone living in the relaxed countryside or the frenetic city?
When we ask people for advice on these fronts we aren’t relying on our advisor for mere facts. We’re asking: given i) what you know about me in particular (how I feel about things, how I react to things, what you see me taking joy in, where my insecurities and fears and anxieties prevent me from doing what I really want to do, etc.) and ii) what you know about life given your experiences, insights, and perspective, what do you think about my options?
But at the end of the day, it’s up to you to decide. That’s what, at least in part, your autonomy consists in. If you’re a good advice taker, you take what they have to say along with everything else you know and think and feel, and you weigh the pros and cons, the values and disvalues, the risks and the rewards, and decide for yourself. This is what it is to act autonomously or self-directed. This is you steering the ship. This is, to use Kierkegaard’s phrase, working out the sum for yourself, even if it isn’t by yourself.
Of course, as an advisee, you are only one half of the advisor/advisee relationship. The advisor, aside from needing the knowledge, experience, and wisdom pertaining to what you’re seeking advice on, must have your best interests at heart. They have to give advice that they think would make you happy/fulfilled/flourish. They must be motivated in the right kind of way to be good advisors.
Three People That Fall Short of This Ideal
This picture of what constitutes an ideal advisor/advisee relationship throws into relief the ways we may fall short of this ideal. I’m going to highlight three ways in the form of three different characters. I don’t mean to say that someone either is or is not one of these characters, though they probably exist. Instead, everyone can be one of these characters at one moment or another.
The Deferrer
Think about a person who seeks advice because they don’t trust themselves to make the decision themselves. They’re insecure. It’s too much responsibility for them. They’re anxious about making the “wrong” choice. “What do you think I should do?” they ask earnestly with furrowed brow.
The deferrer isn’t looking for guidance. They’re looking for instructions. Better, they’re looking for a command. An authority to issue edicts. In doing so, they attempt to absolve themselves of the responsibility of choice. They attempt to forfeit their autonomy, to put the steering wheel in someone else’s hands. Freedom doesn’t fit their psychological temperament.
(There is the complication that any decent existentialist, including Kierkegaard, would note: deferring to another is a choice to defer and so the loss of autonomy is illusory and self-deceptive. But let’s not go down that path).
The Faux Advisee
Some people ask for advice but they aren’t really asking for advice. What they’re really looking for is something like approval or validation of the choice they’ve already made. Or, relatedly, they may (also) be looking for another opportunity to talk about themselves, to make themselves the topic of conversation. For them, advice seeking isn’t an attempt to abrogate their autonomy but is, rather, a form of narcissism.
The Questionable Advisor
In the ideal relationship, the advisor has your well-being as their goal. But some advisors have questionable motives, unbeknownst to the (naïve) advisee. The advisor may, for example, harbor a hidden jealousy of the advisee. Perhaps the advisee already “has more” than them. Or perhaps they would if they went down the path they’re thinking about. Imagine someone considering a career in art or a career in law and consults someone more senior than her who once made a similar decision. We can imagine that person, somewhat regretful of her decision to attend law school, not wanting someone else to take the path she wished she had chosen. And now, armed with the argumentative skills of a lawyer, she can deftly persuade her advisee to avoid pursuing a career in art.
These three kinds of bad advisors/advisees surely aren’t exhaustive of the ways things can go sideways in that kind of relationship and the ideal I’ve sketched can be filled in with many more details. But this is enough to bring out my concern regarding AI advisors.
When the AI Advisor Decreases Our Autonomy, Increases Narcissism, and Offers Questionable Advice
A world where just about everyone has access to an AI advisor has that access 24/7. We immediately see the probability that people will seek the advice of AI far more often than that of humans for the simple reason that it’s logistically easier.
It’s also emotionally easier. ‘What if they think I’m being stupid or ‘just a kid’ or naïve or disloyal or irresponsible?’. Those kinds of thoughts make sense if you have to ask a person for advice. But those emotional impediments simply don’t exist if the advisor is an AI.
What can we expect, then?
If the advice is generally bad then people won’t use it. That’s fine; we’re right where we started before we tried using AI advisors. But if it’s good at advising, or at least better than people’s average advisor, I’d expect people to increasingly become deferrers. Asking AI for lots of advice at increasing frequency. Following its advice more and more. (Think of “automation bias” and people driving into lakes because the GPS told them to). People will increasingly allow AI to steer the wheel. They’ll become habitual deferrers. They’ll allow their lives to be not self-determined, but AI-determined.
I want to highlight that our autonomy is decreased not by virtue of the quality of the advice given. Rather, our autonomy is decreased by “taking the AI’s word for it.” The threat to autonomy is not the quality of the advice but the way in which the advice is received, just as the problem with Kierkegaard’s student isn’t that he has the wrong sum but that he arrived at it in the wrong way.
Meanwhile, while most of humanity busies itself with attenuating what little will it has left, the faux advisee will go full Narcissus as they gaze into the image their AI reflects back to them. And AIs created by companies that want to sell you things will give advice motivated far less by good will and far more by whatever optimizes whatever their business is supposed to optimize for (e.g. time spent with the advisor, clicks on ads, etc.)
AI may well outperform most people in a variety of ways, including as advice givers. But far from entailing that our deliberative work can be outsourced, it means that, insofar as we value autonomy, we should commit to our own deliberative activities all the more.
If you like this article, you’ll probably like my podcast, Ethical Machines.